Susceptibility to SOHIC for linepipe and pressure vessel steels - review of current knowledge

Richard J Pargeter
TWI Ltd

Paper presented at NACE 2007, 11-17 March 2007.

Abstract

Stress orientated Hydrogen Induced cracking (SOHIC) is a manifestation of sulfide stress cracking (SSC) in C-Mn steels by a mechanism involving the linking of planar cracks normal to the primary propagation direction of the crack. It has been known for many years, and the solution has been considered to be the use of hydrogen induced cracking (HIC) resistant steels. Work by the Materials Properties Council has, however indicated that HIC resistant steels may be more susceptible to SOHIC. This raises questions about both the most appropriate steels, and test methods for evaluating them

Published literature has been surveyed, and information pertaining to the mechanisms of SOHIC, with particular reference to the effects of material, fabrication, and service conditions on the risk of cracking, has been extracted. A critical review of this information, along with published information on test methods and results, and information on industrial experience including service failures due to SOHIC, has allowed conclusions to be drawn and recommendations to be made on materials selection, fabrication and testing for resistance to SOHIC.

Introduction

Stress Orientated Hydrogen Induced Cracking (SOHIC) is a manifestation of sulfide stress cracking (SSC) in C-Mn steels, by a mechanism involving the linking of planar cracks normal to the primary propagation direction of the crack. It has been known for many years, and was indeed designated Type I cracking by workers from Sumitomo Metal Industries, ^[1] with the conventional, hardness dependent, SSC being referred to as Type II ( Fig.1). Similarities with hydrogen induced cracking (HIC, also termed Hydrogen Induced Cracking or Stepwise Cracking) were noted, and the solution was considered to be use of HIC resistant steels. ^[1-3] Even at that time, however, there was some uncertainty over the strength of the correlations between HIC resistance, and resistance to SSC. ^[4] More recently, work funded by the Materials Properties Council in USA has indicated that HIC resistant steel may indeed be more susceptible to SOHIC than conventional steels. ^[5] Thus, purchasers are left with some uncertainty over the appropriate steel quality required to provide resistance to SOHIC. Before exploring published evidence on factors affecting SOHIC, however, it is necessary to define the phenomenon, so that earlier papers, which have used different terminology, can be assessed correctly.

Fig.1. Examples of SSC observed in the heat affected zone. A, Type I; B, Type II.

(Ref. ^[1] , ©NACE International 1979)

One factor which has been identified from early on as affecting the risk of SOHIC, is welding. ^[6] Specific attention has therefore been given to papers which have explored the effects of welding on microstructure and properties which could have had a bearing on SOHIC.

Much of the current concern about SOHIC arises from industrial failures, believed to have been by this mechanism. An important part of this review, therefore, has been to obtain as much detail as possible about such failures, with a view to identifying common factors.

In the following Sections, therefore, after considering the definition of SOHIC and clarifying steel terminology, the fundamental mechanism of SOHIC and microstructural effects are explored before discussing the effects of welding and post weld heat treatment. Test methods and service failures are addressed separately.

Definition of SOHIC

Morphologically, there is agreement in the literature that SOHIC consists of arrays of small cracks in the rolling plane, aligned normal to the causative stress (usually in the plane of the plate), and linked by cracks normal to the stress ( Fig.2). There is clearly a close link between this and the stepwise form of hydrogen induced cracking (SWC), but the alignment of the cracks in a stack, normal to the stress (applied and/or residual) is different from the stepped morphology in the absence of stress. The distinction between SOHIC and SWC, however, is not precise, and indeed both can occur simultaneously in a susceptible steel ( Fig.3). This causes a problem in the collation of data from both testing and service experience; should failures attributed to SWC be included or not?

Fig.2. Schematic illustration of the morphology of hydrogen-induced stepwise cracks, SSC-I, and SSC-II. (Ref. ^[46] , ©NACE International 1993)

Fig.3. SOHIC and HIC in different regions of a bend specimen - tensile surface at the top

For the purpose of this review, the emphasis will be very clearly on SOHIC in steels which are not highly susceptible to SWC. It is evident that through thickness cracking can occur in steels which are susceptible to SWC; the problem is identifying the factors contributing to SOHIC in other steels, steels in which HIC needs to be 'oriented' in order to produce through thickness cracking, and indeed steels in which HIC is not easily initiated.

Thus, SOHIC is defined as cracking in rolled or forged steel, normal to the rolling or forging plane of the steel, which consists of linked stacks of small cracks lying in the rolling or forging plane of the steel.

Steel terminology

There has been a certain amount of controversy over the effects of steel type and quality on the risk of SOHIC, and much of this is due to differences in terminology, and in particular different perceptions of what constitutes a'clean' and a 'sour service' steel. In work carried out for the Materials Properties Council, ^[5] four types of steel were defined, as follows:

Conventional Steel (CS)

A conventional steel is a commercially produced steel which is either hot rolled or normalized (e.g., ASTM A285C or A516-70). It has generally moderate to high levels of impurities, particularly sulfur (i.e. ≥0.010 wt% sulfur). This type of material generally has a high susceptibility to HIC in most hydrogen charging environments even under moderate exposure conditions.

Low Sulfur Conventional Steel (LSCS)

A low sulfur conventional steel is a commercially produced steel (e.g., ASTM A516-70) that contains lower than normal levels of sulfur (i.e., 0.003 to 0.010 wt%). This steel can exhibit improved mechanical properties over conventional steels, but typically has not been processed to specifically exhibit high resistance to HIC. These steels can still show significantly high susceptibility to HIC even in moderate service environments.

HIC-Resistant Steel (HRS)

The term 'HIC-resistant' steel is used by manufacturers and users to denote conventional grades of steel (e.g., ASTM A516-70) that have been metallurgically processed to enhance their resistance to HIC. Such processing typically includes ultra-low sulfur levels (i.e. ≤0.002 wt%), normalizing heat treatments to modify the hot-rolled microstructure and possibly calcium additions to provide sulfide shape control. These steels typically have improved resistance to HIC as compared to conventional steels, however, they may still show some degree of susceptibility to HIC and SOHIC in severe wet H2S service conditions.

Ultra-Low Sulfur Advanced Steels (AS)

Ultra-low sulfur advanced steels are those made by the most modern steel making and processing techniques. These steels typically have ultra-low levels of sulfur (e.g. ≤0.002 wt%) and low carbon equivalents compared to conventional steels of comparable tensile strengths (i.e., ASTM A516-70). Steels in this category are currently made to ASTM A841 by TMCP (Thermo Mechanical Controlled Processing) and/or accelerated cooling techniques. Also, they have reduced carbon levels as compared to conventional steels to produce ferritic or ferritic/bainitic microstructures with little or no microstructural banding.

Whilst this may represent typical North American practice, however, this is not representative of current European terminology, certainly for corrosive oil and gas service. 'Conventional' European steel in this context would typically have around 0.005-0.008%S, this being to ensure good through thickness and general toughness properties, i.e. the LSCS above. In European practice, a 'low sulphur' steel would probably have ≤0.008%S, this being to ensure good thorough thickness and general toughness properties. A linepipe steel designed for sour service would typically have ≤0.003%S, controlled Mn and P levels, Ca treatment and a carbon level between ~0.04 and 0.08%, depending on the steel processing; higher carbon can be tolerated in an accelerated controlled TMCP or QT steel than in a conventional controlled rolled or normalised steel. This composition fits with the definition of 'HIC resistant steel' above, but more controls on processing than just normalising would be required to ensure resistance to HIC. (The significance of processing route is recognised in both EFC16 and ISO 15156-2 ^{[7, 8]} ). In particular, segregation (during casting) needs to be minimised, and heavily banded ferrite/pearlite structures are known to be at risk. In pipeline manufacture the benefits of TMCP, particularly with accelerated cooling, have been recognised for many years, and for thicker (pressure vessel) plates, quenched and tempered steels have been claimed to give improved HIC resistance, ^[9] although with careful production and tempering, normalised HIC resistant pressure vessel plates can also be produced. ^[10]

Thus, for the purposes of this review, the following definitions will be used:

Low Sulphur Steel:≤0.008% Very Low Sulphur Steel:≤0.003% HIC Resistant Steel:Steel with an appropriate chemical composition and processing route, which gives CLR, CSR and CTR values of ≤15%, ≤3% and ≤1.5% respectively on NACE TM0284 tests in solution A (pH3).

Conditions required for the occurrence of SOHIC

Leaving aside materials issues, SOHIC requires an environment which introduces hydrogen into the steels, and stress. It is a common observation that both severe hydrogen charging conditions and high stresses are required for SOHICto occur, at least in steels which are resistant to HIC. ^[11] Crolet ^[12] states that this combination is needed 'to initiate the local HIC required to induce SOHIC', although specific conditions are not defined. Schmitt et al ^[13] comment that 'Obviously, 60% YS is enough to produce SOHIC under critical conditions'. Christensen ^[14] points out that 'The few cases [of SOHIC failures] that are reported in public strongly point to the necessity of both a very aggressive environment.... and localised plastic straining'.

As will be discussed further under 'fundamental mechanisms' below, it is believed that initiation of small HIC cracks is a necessary step in the formation of SOHIC. Stress can contribute to this, and particularly triaxial stresses. These can develop in inhomogeneous regions, such as are associated with welds and also in complicated connections such as nozzle joints. Very high initial hydrogen fluxes, such as may occur when a clean steel surface is first exposed to a sour environment, may also facilitate this first stage. It has been found in the MPC programme ^[15] that short, extremely intense hydrogen charging can result in more cracking than longer exposure to an environment more severe than the background level, but less severe than the peak intense hydrogen charging condition.

Fundamental mechanism of SOHIC

The definition of SOHIC involves linked stacks of small cracks lying in the rolling (or forging) plane of the steel. Thus, an understanding of the mechanism of cracking requires an understanding of the formation of both the in-plane cracks, and the linking cracks, and the interactions between them. The interaction between the cracks also needs to explain the orientation of in-plane cracks into stacked arrays.

Starting with the formation of in-plane cracks, it is generally agreed that these consist of HIC. ^{[16, 12]} Nevertheless, SOHIC can occur in materials which are resistant to HIC. In these cases, HIC is apparently initiating on interfaces which are stronger than inclusion: matrix interfaces (where HIC is generally found in unstressed specimens) such as at pearlite colonies. ^[17] For this to occur, severe stress and/or hydrogen charging conditions are necessary. To some extent these two factors are linked, as a high hydrostatic stress increases the solubility of hydrogen in the steel. Thus, even a principal applied or residual stress in the rolling plane can encourage the nucleation of HIC, by increasing local hydrogen levels. In inhomogeneous materials, however, such as welds, variations in strength can lead to the development of high through thickness stress, which would directly encourage the formation of in-plane cracks. Takahashi et al ^[18] have shown by finite element analysis that high through thickness stresses can develop at mid wall in a softened (HAZ) region under the influence of transverse loading, due to strain concentrations.

Once an initial HIC has formed, the tips of the crack are subject to stress from two primary sources, namely external, (residual and/or applied), stresses, and stresses due to hydrogen pressure within the crack. It has been shown both analytically and experimentally that internal pressure alone can cause through thickness linking of cracks, resulting in stepwise cracking (SWC). ^{[19, 20]} The work by Iino, ^[19] however, has explored the combination effects of external and internal (hydrogen pressure) stresses. He demonstrated that the shear stresses imposed by an externally applied (in-plane) stress actually oppose those generated by pressure within the crack thereby stifling stepwise linking. As the applied stress increases, the location of maximum shear stress moves from very close to the crack tip to mid way along the crack, and a little way from the crack inthe through thickness direction. This effect is demonstrated in Fig.4. Iino proposed that the principal significance of the shear stress was to encourage the nucleation of cracks at (e.g.) inclusion/matrix interfaces by dislocation pile-ups, whilst also recognising the importance of hydrostatic tensile stress in encouraging hydrogen diffusion into the material around the initial crack. What is also evident from Fig.4, is that for the same peak level of shear stress, a lower internal pressure is required in the presence of an externally applied load. Thus, earlier cracking, in a stacked array, might be expected in a stressed material. Iino also demonstrated experimentally that under high enough applied loads, cracks were aligned on, and linked along, a 45° shear plane, which remains consistent with his theory on the importance of shear stresses.

Fig.4. Combined effects of shear stresses from internal pressurisation and in-plane stressing on position of maximum shear stress

(Ref. ^[19] , Reprinted with permission of ASM International ^®. All rights reserved. www.asminternational.org)

Crolet ^[12] has proposed that in-plane cracking must form first, and then link through thickness, as no through thickness stress can exist after the linking has occurred. The 'Latent initiation site' model proposed by Amano et al ^[17] explains how HIC cracking can be encouraged within the plastic zone associated with a pre-existing HIC through hydrogen transport along an activated slip plane ( Fig.5). The higher the applied stress, the greater the extent of the plastic zone in the through thickness direction, and therefore the greater the chance of intersecting a latent initiation site ( Fig.6). This is consistent with the work of Iino, reported above, which indicated that maximum shear stresses rotated around pre-existing cracks as the external stress increased. On the basis of these theories, and as evidenced by numerous examples of samples containing stacks of in-plane cracks which have not linked to cause fracture, it is clear that the in-plane cracks do form first, but it is also likely that through thickness linking can occurin more highly stressed regions, while in-plane cracking is still occurring ahead of the macroscopic crack tip. The overall sequence was sketched by Ohki ^[3] and is reproduced in Fig.7.

Fig.5. Schematic explanation of latent initiation site model. Conditions for propagation of SOHIC:

λ ₁ < λ _c : SOHIC propagation is promoted.

λ ₂ < λ _c : SOHIC propagation is not promoted.

(Ref. ^[17] , reproduced by permission of MPC)

Fig.6. Local yielded region; stressed condition (for SOHIC) expands yielded region in perpendicular direction to stress

(Ref. ^[17] , reproduced by permission of MPC)

Fig.7. Schematic of crack propagation mechanism in SOHIC:

1) Microcracks form under the notch bottom and start to grow;
2) The pit formed at the notch bottom connects to the microcracks ahead;
3) Microcracks form ahead of the main crack with MnS inclusions as nuclei;
4) The main crack connects to these and propagates
(Ref. ^[3] , Reprinted, with permission, copyright ASTM International)

Crack linking is most probably by a slip rather than cleavage or intergranular mechanism. It was reported by Miyoshi et al in 1975 ^[21] that hydrogen cracking in a single crystal of pure iron occurred on a slip plane. Furthermore, the significance of hydrogen transport along activated slip planes has been highlighted above. Several authors have reported transgranular cracking, but this is most clearly shown in Fig.8, taken from Ref ^[22] . A 'quasi cleavage' fracture around an oxide cluster is reported by Takahishi et al ^[18] , but it is not entirely clear from the evidence presented whether this is the crack morphology, or caused by breaking open after chilling in liquid nitrogen. (It should be recognised, however, that quasi cleavage fracture can occur at around 200HV in weld metal fabrication hydrogen cracking). No reports of intergranular fracture associated with SOHIC have been found.

Fig.8. Transgranular crack (HSSC) of a SOHIC, starting at the stressed surface

(Ref. ^[22] , ©NACE International 1985)

Microstructural effects

Despite the apparent similarities between HIC and SOHIC, the susceptibility to SOHIC does not appear to be dominated by inclusions in the same way as HIC. Indeed, some authors have claimed that 'HIC resistant' steels are more susceptible to SOHIC than 'conventional' steels. To some extent, this may be an artefact of the testing, or a mis-reading of test results. The data generated by Iino, ^[23] for example, shown in Fig.9, have been cited as evidence that reduced sulphur levels increase susceptibility to cracking under severe hydrogen charging conditions. In fact, this figure gives no indication of the effect of sulphur on threshold stress, but solely on the rate of cracking, and Iino showed a correlation between time to failure, and time to attain a given number of diffusible hydrogen atoms per inclusion. (Clearly, the 0.001%S steel which did not crack does not follow this trend, the correlation only being possible for steels which crack).

Fig.9. Results of constant load test under hydrogenation (8mA/cm ² ).

(Ref. ^{[23, 5]} , ©NACE International 2002)

Christensen ^[14] has warned against the potentially confusing effects of traps on time taken for crack initiation. He argues that, in the presence of a large number of trap sites (e.g. a high sulphur steel) the build up of hydrogen in the plastic zone associated with a notch in a test piece may be delayed. If this delay is sufficient, a lower peak hydrogen level may be achieved in the plastic zone than in a clean steel, as there is a marked reduction in hydrogen charging rate as the initially clean steel surface becomes scaled. Thus, dirtier steels may show a higher threshold stress for SOHIC in testing because of the beneficial effect of hydrogen traps during the initial transient stages of hydrogen charging. Depending on the surface condition of steels in service, such initial surges in hydrogen generation may not occur, and this effect of steel (internal) cleanliness may not be observed.

Christensen also points out that if a crack, propagating through wall, intersects with a planar HIC, then it will be blunted, and possible arrested. While this could occur in a bend test, where there is a through thickness stress gradient, the argument is not consistent with the observed mechanism discussed in Section 4 above, whereby the first step is HIC formation, followed by through thickness linking.

Although much of the reported good behaviour of older, dirtier steels may be an artefact of trapping effects, it is nevertheless evident that there are other important microstructural factors affecting the behaviour of cleaner steels. Several authors have reported that pearlite colonies can act as sites for HIC ^{[17, 24, 25]} and Coudreuse et al ^[11] have demonstrated how a quench and temper heat treatment can significantly reduce or eliminate cracking in a TM0284 test on A516 grade 60 normalised steel. Similarly Amano et al ^[17] improved the SOHIC resistance of parent plate by 'dispersing pearlite colonies' through thermomechanically controlled processing (TMCP). Banded structures have been identified as at risk from SOHIC by Cayard et al ^[5] , Blondeau ^[26] and Condreuse et al. ^[11]

Cayard et al ^[5] have produced an equation for Predicted Average Cracking Ratio in stressed four point bend specimens, which incorporates a banding index (measured according to ASTM E1268) but also whether the steel is Ca treated and the carbidephase microhardness:

PACR = 35 F _Ca BI+0.05(HV _12.5g - 330)

Where:

PACR = predicted average cracking ratio
F _Ca = calcium treatment factor (Yes = 1; No = 0.1)
BI = banding index in accordance with ASTM E1268
HV _12.5g = Vickers hardness using 12.5 gram load.

They commented earlier ^[27] that banding became an important variable 'with reduced sulfur content'. It is not clear over what range of sulphur contents this equation is applicable.

Crack initiation has also been attributed to other microphases. ^{[24, 28]} For example, Gianetto et al ^[28] presented clear micrographs showing cracked martensite/austenite (MA) microphases in weld HAZs, but the overall cracking in these cases did not have a SOHIC morphology. Overall, although the possibility of martensitic islandsinitiating SOHIC within soft HAZs has been considered by several authors, this possibility has generally been discounted. For example, Takahashi et al ^[18] reported that there was no evidence for martensite, either optically or from hardness measurements, and concluded that 'the formation of martensite is not sufficient to account for the cracking in the soft HAZ'. Cialone and Williams ^[29] also discounted an effect of martensite/austenite constituents after PWHT failed to reduce sensitivity to cracking, and in view of the fact that no direct interactions were observed. Ume et al ^[25] also explored the effect of MA in intercritical regions, but found no evidence for such constituents. They did, however, find evidence of crack nucleation on retained pearlite in this location.

It should be recognised that there are also other microstructural features, other than non metallic inclusion, which contribute to trapping, hydrogen solubility and hydrogen diffusion in steels. Blondeau ^[26] identified the importance of alloy carbides (V, Mo, Nb, Ti, Zr) and also commented that steel rolled in the two phase region would have more traps than that rolled as austenite due to strain effects.

Only one report has been found of investigation of cracking by transmission electron microscopy. Cialone and Williams ^[29] found no differences in dislocation structure in crack sensitive and insensitive regions of welded samples, but did find that crack sensitive regions consistently contained relatively high concentrations of 100-200Ådiameter Nb rich precipitates. High concentrations of ≤50 Å precipitates, or a few >500 Å did not apparently affect sensitivity. Although no mechanism was postulated, they attributed the poorer behaviour of Nb microalloyed steel to this microstructural effect. Ume et al ^[25] found intergranular cracking close to the fusion boundary of a 0.02%C SiMnNbB steel, which they attributed to a combination of carbon depletion and Nb(CN) precipitation weakening grain boundaries, while Nb (CN) precipitation also strengthened the matrix. Although this was not a SOHIC mechanism, there may be some parallels. By contrast, in work by Kobayashi et al ^[30] it was only the Nb free steel, out of the three welded steels tested, which cracked.

Effects of welding

It is commonly observed ^{[5, 17, 18]} that the outer, commonly softer, regions of weld HAZs are particularly susceptible to SOHIC, indeed SOHIC in this region has been termed 'Soft Zone Cracking' (SZC). Cayard and Kane ^[5] report a 'decrease in threshold stress of about 10 to 20 percent when compared to the corresponding base metals', although according to the data presented, there was no change for the two 'HIC-resistant steels' tested (see Section 4). They reported HAZ hardness of 'less than HRB 100, converted from 500 gram Vickers readings' (~240HV, converted back using ASTM E-140). No indication of the location of cracking within the HAZ was given, however.

Amano et al ^[17] reported SOHIC in the HAZ of a 15mm thick API 5L X46 controlled rolled steel, with 0.07%C, 0.93%Mn, 0.005%P, 0.008%S, 0.030%Al microalloyed and Ca treated, and welded in two passes at 5kJ/mm. This was tested using NACE TM0177tensile specimens and they report that: 'The SOHIC occurred in the temperature range 700-800ºC, which was just above the AC1 temperature and off the most softened zone'. They go on to say that 'The fact that the SOHIC location was not in the most softened zone, indicates that the SOHIC occurrence is affected not only by the lowest yield strength and/or the susceptibility of the microstructure but also by the localised stress and/or strain elevation associated with heterogeneous plastic deformation'. They then proceed to develop their 'latent initiation site' model, in which in-plane cracks are encouraged to form at sites such as pearlite colonies within the yielded region around an initial HIC ( Fig.5 - see Section 5). Their finite element analysis demonstrates that the highest axial stress, near the surface is just outside the softest region of the HAZ ( Fig.10), which explains their observed cracking in this location. They do warn, however, against the risk of cracking initiating at positions of higher stress within the specimen (which also coincide with lower strength and greater material susceptibility to SOHIC) in a less clean steel 'in which SOHIC initiates preferentially from the inside associated with high density of non metallic inclusions, for example'. It is apparent from the data plotted in Fig.10 that they have assessed the susceptibility of different HAZ microstructures, presumably using simulated HAZ specimens, and found some variations in an X65 steel. No microstructural or other explanation for this is given, although this variation appears to follow the reported variation in strength (measured on simulated specimens, Fig.11) to some degree.

Fig.10. Local axial stresses in round bar specimen including HAZ and SOHIC susceptibility in HAZ; region where local stress exceeds susceptibility is in high risk of failing in SSC test.
(Ref. ^[17] , reproduced by permission of MPC)

Fig.11. Simplified yield strength model with dependence on peak temperatures

(Ref. ^[17] , reproduced by permission of MPC)

The softened zone was also explored by finite element analysis by Takahishi et al ^[18] , and they found that through thickness stresses peaked within the softest region, at mid thickness, under the influence of a cross weld applied load, due to strain concentration ( Fig.12). This would encourage the formation of initial in-plane cracks in this region.

Fig.12. FEM analysis of a weld including a softened region in the middle of harder regions:

a) FEM model;

b) Stress distribution derived from FEM analysis.
(Ref. ^[18] , reproduced by permission of Gulf Publishing Company)

Effects of post weld heat treatment

In work for the Materials Properties Council ^[5] Cayard et al have reported a general decrease in threshold stress for SOHIC (measured using NACE TM0177 tensile tests) in welded samples following PWHT. In one 'HIC resistant steel' (0.15%C, 0.001%S, 1.15%Mn, A516-70) the threshold was reported to be very close to the design value of 25% specified minimum tensile strength permitted by ASME VIII division 1. It was suggested that the reduced threshold could be due to a loss of strength, possibly having a direct effect on susceptibility, and also slightly increasing the stress in terms of percentage of actual tensile strength.

No other references to PWHT increasing the risk of SOHIC have been found, but Cialone and Williams ^[29] reported no effect of PWHT. By contrast work carried out by Hoesch Rohr, ^[31] comparing on line PWHT spirally welded pipe with longitudinally seam welded pipe found a clear beneficial effect of reduced residual stress levels on the occurrence of SOHIC. Thus, it is possible that PWHT increases the inherent susceptibility of the HAZ (or parent material) microstructure to SOHIC, as suggested by the small scale tests in Ref ^[5] , but the benefit of stress reduction, as observed in the full ring tests used in Ref ^[31] , can override this effect. This possibility is, indeed acknowledged in Ref ^[5] . This would not be the explanation for the improvements in resistance to SOHIC on PWHT recorded by Coudreuse et al using the double beam four point bend test on a selection of quenched and tempered steels. ^[11] With this test, residual stress effects would not be expected, and any effects would be expected to be material effects. SOHIC was eliminated in all steels tested, including the 'standard' material, which experienced 38% TCL(Total Crack Length) in the as welded condition at 80% SMYS. Regardless of a possible benefit of PWHT through stress relief, however, Merrick and Bullen ^[32] took the view that, with threshold stresses for SOHIC at 30% -50% yield, the benefits of thermal stress relief were insufficient to be relied upon. Nevertheless, they point out that prevention of SSC in initially hard regions(through tempering) can prevent a stress concentration forming, which could extend by a SOHIC mechanism.

Thus, there is general agreement that reductions in residual stress resulting from PWHT are beneficial with regards to SOHIC, but it is possible, at least in some steels, that there is a contrary adverse effect on material susceptibility, and in any case, the degree of stress relaxation due to PWHT may be insufficient to be of practical use.

Test methods

In a paper by Merrick and Bullen in 1989 ^[32] , it was reported that 'steels that are resistant to stepwise cracking are also resistant to SOHIC. Therefore TM-02-84 can be considered as a qualification test for steels to resist both HIC and SOHIC. The tests conducted todate suggest a CLR value less than 15% as an acceptance level for pressure vessel steels to be SOHIC resistant since no steel with a low CLR value experienced SOHIC'.

This conclusion was based on NACE TM0177 tensile tests, sectioned after exposure to reveal any stacked arrays of cracks. Tests were carried out at 30%, 50% and 90% yield and steels which did not crack at 50% yield were considered resistant to SOHIC.

The NACE tensile test is included in the 'NACE Standard Test Method for Laboratory Test Procedures for Evaluation of SOHIC Resistance to Plate Steels Used in Wet H2S Service'. ^[33] The standard suggests that the basis of test evaluation should be time to failure, but includes options for sectioning and metallographic examination, or pulling to failure after both completion of the exposure period and subsequent hydrogen release, to determine the effect of any internal cracking on the load bearing capability of the specimen. The principal technique presented in this test method, however, is the double beam, four point bend test. This has been developed by MRC and in its final form consists of two samples, each containing a central electrodischarge machined notch, bolted back to back across a pair of rollers, thereby putting both into four point bending ( Fig.13). A comparison between these two test methods was made by Coudreuse et al ^[11] , using quenched and tempered pressure vessel steels. They found a marked difference in performance of one of their steels, which showed no SOHIC in tensile specimens which had survived 720 hours at yield (~500MPa) but which showed significant cracking in the double beam test, even at 300MPa. They explain this in terms of a difference in SOHIC mechanism between QT and normalised steels. Indeed, they claim that the QT steels they were testing did not display classical SOHIC cracking (ladder like arrays), and that there was no HIC component. (No micrograph was presented to allow an independent judgement on this). They state that, in a QT steel 'SOHIC is effectively equivalent toSSC', and because this propagates rapidly, only one of the double beams fails, making quantitative results problematic. They concluded that the tensile test is more suitable for QT steels, whereas the double beam test is successful when applied to normalised steels, but overall, they preferred tensile testing for all materials. They did not carry out tests for SOHIC resistance on any welded samples.

Fig.13. Double beam test specimens from NACE TM0103-2003
(Ref. ^[33] , ©NACE International 2003)

Siegmund et al [34] compared plain 4 point bend, notched 4 point bend, and the double beam 4 point bend tests. Although notching clearly helped to induce SOHIC, they favoured plain 4 point bend specimens for material screening purpose. The reason given was that the stress state, although triaxial, was 'completely undefined', and the double beam approach was ruled out because it was considered to be much less well defined.

It is observed that SOHIC tends to occur in situations where there is high triaxial stress, such as around nozzles in thick pressure vessels and in spiral welded pipe in ring ovalisation. For this reason, large scale tests are sometimes favoured. It has, for example, been reported by Fowler ^[35] that one of the main reasons behind the development of the full ring test ^[36] was the ability to test for susceptibility to SOHIC. One feature of this test, as performed by Fowler, and as defined in OTI 95 635 ^[36] is that the internal surfaces of the pipe are prepared by grit blasting. This results in a very high initial hydrogen entry rate into the steel, ^[37] and may encourage SOHIC initiation. It was stated by Fowler ^[36] that the ring test procedure had been validated by comparison with sour pressure tests, but the internal surface preparation was presumably the same for both types of test.

A comparison was made by Takahishi et al ^[38] between full ring testing on an X65 UOE pipe and small scale tests. Whereas no SOHIC was seen in the ring test, or single sided exposure rectangular tensile specimens, SOHIC was induced in softened HAZ regions in fully immersed round and rectangular tensile specimens at applied stresses of 0.725 actual yield stress and above. The authors attributed this difference in behaviour to differences in stress state (they demonstrated multi axial stress levels within the softened HAZ in tensile tests by FE analysis) and differences in hydrogen entry (single sided exposure vs. full immersion). They concluded that a single sided exposure, rectangular tensile test is an appropriate small scale test technique, which showed good agreement with the full ring test.

One important feature of any test method for SOHIC, is that it should provide a measure of crack initiation, rather than crack arrest. It is argued by Christensen ^[14] that the NACE TM0103-2004 double beam test is so severe (yielding notch tip, plus very rapid initial hydrogen charging) that, particularly in clean materials, crack initiation is enforced in many essentially resistant materials, and the test measures crack arrest. He proposes that by step wise increase of the test environment severity, the very severe initial hydrogen surge can be avoided, and threshold conditions for SOHIC initiation can be determined.

Industrial experience

An attempt has been made to collate information on instances of SOHIC which have occurred in industry. In doing this, only failures in which there is a fair degree of certainty that the primary mechanism was SOHIC have been included. Even so, the available information has not always been sufficient to make an independent judgement, and reported diagnoses have had to be accepted in some cases. Information has been sought from companies through an EFC/EPERC survey and by direct request to selected companies, and published literature has also been searched. A summary of the instances which have been identified is given in Table 1.

Table 1 : Summary of industrial SOHIC failures

Of the 10 examples in Table 1, only two are downstream pressure vessels. The remainder were all pipelines, four spiral welded, two seamless and two ERW. In the two seamless pipes, and in the amine absorber tower ^[42-44] , the SOHIC propagated from cracks caused by sour service, but via another mechanism. In both the seamless pipes the initiating feature was a blister in the parent material, reported in one case ^[42] to have formed at alumina/iron oxide/slag inclusions, remaining due to inadequate ingot cropping, and the mechanism was judged by the author to be stepwise cracking (SWC) rather than true SOHIC. In the amine absorber tower ^[43] the cracking had developed from sulfide stress cracks in hard HAZs, by a mixture of SWC and SOHIC. In one of the two ERW pipe failures, micrographs presented showed SWC in the re-oriented microstructure. All other failures were judged to have been caused by true SOHIC.

Considering just the true SOHIC failures, in all but two cases there is evidence to show that the steels would not be defined as even 'low sulphur' (see Section 4 above), with reported sulphur contents ranging from 0.01 to 0.025%.In two of those cases, ^{[39, 41]} HIC test results were also presented and these were well outside recommended limits for sour service. In one of the two remaining cases neither chemical composition nor HIC test data were reported. In the final case, ^[40] the material would just fit the 'low sulphur' designation, and no HIC test data were presented. Thus, no examples of failures in very low sulphur or HIC resistant steels have been found. The possible exception to this is the Saudi Aramco 1998 failure, ^[45] where the pipe material was subjected to HIC testing in Solution A, and 'passed'. Although the sulphur content was reported as 0.011%, there is a footnote to the table stating that the S & P results were 'questionable'. Inthis case, however, clusters of CaAlO inclusions were found associated with the cracking.

Although none of the failed materials would have been expected to be strongly resistant to SOHIC, it is evident either from the reported H ₂ S and CO ₂ contents and/or evidence of corrosion damage, that the environmental conditions were very severe in all cases. Less complete information is available on residual and service stresses. In one case, ^[45] residual stresses were very high, but it was not clear that this was true in other examples.

The overall message from the review of failures is that these have all occurred in material which would be expected to have sites for initial HIC formation, under the influence of a severe hydrogen charging environment. The picture with regard to stressing is less clear, but there is no conflict with the general perception that high service and/or residual stresses are required for SOHIC to occur.

Discussion

Mechanism of cracking

The similarities and features in common between hydrogen induced cracking (HIC), particularly in the form of stepwise cracking, and SOHIC, has led to confusion both in understanding the mechanism of cracking and its control. Reduction or elimination of rolled inclusions in steel (a primary control for HIC) is certainly necessary to prevent cracking in stressed steel under severe hydrogen charging conditions, but may not be sufficient. SOHIC is defined inthis review as cracking in rolled or forged steel, normal to the rolling or forging plane of the steel, which consists of linked stacks of small cracks lying in the rolling or forging plane of the steel. Theoretical analysis by Iino ^[19] has demonstrated how the interaction of stress fields is necessary for the formation of stacks of in-plane cracks, but there is evidence that the in-plane cracks form before the through thickness components. Although this would apparently indicate the need for a significant microstructural weakness, such as an inclusion interface, both experimental observation and theoretical analysis have shown that through thickness stresses can develop under the influence of axial loading, sufficient to induce in-plane cracking in certain microstructures and in particular, in banded ferrite: pearlite structures. Such through thickness stresses are particularly able to form in locally softened material,such as outer HAZ regions through strain concentration. It is possible that the observed effect of banding, and relative pearlite/ferrite hardness, is also due to strain concentration on a small scale, although it may be also due to apearlite:ferrite interface effect. In steels which have been processed to ensure good microstructural homogeneity (e.g. quenched and tempered steels) less homogeneous and more susceptible microstructures are also likely to develop in such softened regions.

Considering again the initiation of in-plane cracks, a contributing factor which has been identified is diffusible hydrogen content. Under conditions of long term steady state exposure, the diffusible hydrogen content of the steel will be related directly to the corrosion rate, but the initial build up of interstitial diffusible hydrogen within the steel will also be affected by the presence of traps within the steel - a large number of traps will absorb some ofthe input hydrogen and slow the build up of diffusible hydrogen. In general, clean steels contain fewer traps. This effect has implications for non steady state situations and particularly for initial exposure. Very high hydrogen fluxes are seen when a clean steel surface is first exposed to a sour, acid, environment but the flux reduces quite rapidly as the clean surface becomes contaminated. Thus, the diffusible hydrogen concentration in the steel will initially rise and then decrease as steady state conditions are achieved. If the steel has large numbers of traps, the height of the concentration peak may be reduced, by comparison with that for a cleaner steel, resulting in a greater driving force for SOHIC initiation in the cleaner steel. The significance of such an effect depends on the relative conditions in testing and in service. In many service conditions, steel surfaces will not be clean when first exposed to severe conditions. Thus, a particularly high peak concentration, early in a test, which may initiate SOHIC, is unrepresentative and furthermore may unreasonably penalise clean steels, which would be expected to have greater inherent resistance to SOHIC under steady state conditions (for similar microstructures, strength etc). It must not be forgotten, however, that there are service conditions (hydrofluoric acid, for example) in which very high initial hydrogen fluxes are experienced and thus, the anticipated service conditions need to be taken into account when devising test conditions.

Testing and control

The above understanding has implications for control measures which may be taken to prevent SOHIC, starting with material testing and evaluation. It is evident, for example, that the resistance of a weld needs to be evaluated and that the test method needs to take into account both microstructural and local strength changes.

There are several test methods which have been used to evaluate the risk of SOHIC, of varying degrees of complexity and severity. It is evident that SOHIC can be induced in simply stressed specimens under sufficiently high applied loads and such testing may be useful for material ranking. For prediction of service behaviour, however, both an appropriate tri-axial stress state and appropriate hydrogen charging conditions through the test period are required. Any test method also needs to be applicable to welded samples. No such small scale test method currently exists, although there are some developments underway. ^[47]

With regard to guidance on material selection and fabrication, despite the deficiencies in current test methods, some useful comments can be made. It is likely that the observations in some work of poorer performance of clean steels is an artefact of the test conditions, and in particular an initial surge in hydrogen charging. With the exception of service conditions (such as HF) where a similar initial surge in hydrogen charging may be expected, it is considered that steels should be as clean as possible. Depending on the duration and severity of the initial hydrogen surge in ( eg) HF service, protection from trapping effects cannot be guaranteed to be successful in any case. There is no disagreement that microstructural homogeneity is beneficial, and in this respect, heavily banded ferrite:pearlitesteels are at greatest risk, while quenched and tempered and TMCP steels with accelerated cooling are most resistant. Both QT and TMCP steels with accelerated cooling however, may suffer both softening and microstructural changes in the HAZ which increase the risk of SOHIC in this location. High heat input welding would be therefore expected to be detrimental, particularly in such materials. It is not clear whether PWHT is detrimental or beneficial to materialsusceptibility, but the stress relieving effects are apparently beneficial, and to a sufficient extent to ensure that the overall effect of PWHT on risk of SOHIC is beneficial.

Conclusions

From a review of published and other information on stress oriented hydrogen induced cracking (SOHIC), the following conclusions have been drawn.

1. Experimental work indicates that both a severe hydrogen charging environment and high stresses are required for SOHIC to occur, at least in HIC resistant steels. This is largely supported by industrial experience.
   
   
2. Short periods of intense hydrogen charging may sensitise a steel to SOHIC, by initiating small HIC cracks. Very clean steels, with few hydrogen traps, may be at particular risk from this effect, although inherently resistant to SOHIC.
   
   
3. It is considered likely that complex, triaxial loading, can encourage formation of small HIC cracks by stressing normal to the rolling plane, and by increasing hydrogen solubility in the steel.
   
   
4. Reduction or elimination of rolled inclusions is necessary to prevent SOHIC in stressed steel under severe hydrogen charging conditions, but may not be sufficient.
   
   
5. Microstructural inhomogeneities, such as ferrite:pearlite interfaces, particularly in heavily banded steel, can provide initiation sites for SOHIC.
   
   
6. Outer HAZ regions, particularly in high heat input welds, have an increased susceptibility to SOHIC both because of strength and thus strain inhomogeneity, and the development of less homogeneous microstructures in certain ( eg quenched and tempered) steels.
   
   
7. The effect of PWHT on material susceptibility to SOHIC is not clear, but the overall effect on risk of SOHIC is likely to be beneficial.
   
   
8. It is important that test methods for SOHIC accurately represent the combination of triaxial stressing and hydrogen charging that may occur in service, and are applicable to welded material. No such small scale test method currently exists.
   

References

1. Kaneko T, Takeyama M, Nakanishi M, Sumitomo Y and Ikeda A: 'Improvements of hydrogen sulphide cracking susceptibility in line pipes for sour gas service.' First NACE Middle East Corrosion Conference, Bahrain, 15-17 April 1979 pp932-954.
   
   
2. Ikeda A, Monita Y, Terasaki F and Takeyama M: 'On the Hydrogen Induced Cracking of line pipe steel under wet hydrogen sulphide environment'. Second International Congress on Hydrogen in Metals, Paris, France 6-11 June 1977. Paper 4A7, pp1-8. Pergamon Press, Oxford 1978.
   
   
3. Ohki T, Tanimura M, Kinoshita K and Tenmyo G: 'Effect of inclusions on Sulphide Stress Cracking'. Symposium on Stress Corrosion - New Approaches, Montreal, Canada, 22-27 June 1975. ASTM Special Technical Publication 610, pp 399-419.
   
   
4. Herbsleb G, Popperling R K and Schenk W: 'Occurrence and prevention of Hydrogen Induced Stepwise Cracking and Stress Corrosion Cracking of low allow pipeline steels'. NACE Corrosion/80, Chicago, 3-7 March 1980. Paper 9 and Corrosion Vol.37, No 5 May 1981, p247-256.
   
   
5. Cayard M S, Kane R D and Horvath R J: 'SOHIC resistance of C-Mn plate steels used in refinery service'. NACE Corrosion/2002, Denver, Colorado, USA, 7-11 April 2002. Paper 02554.
   
   
6. Ikeda A, Terasaki F, Takeyama M, Takeuchi I and Nara Y: 'Hydrogen Induced Cracking (HIC) susceptibility of various steel line pipes in the wet sulphide environment'. NACE Corrosion/78, Houston, Texas, USA. 6-10 March 1978. Paper 43.
   
   
7. European Federation of Corrosion Publications No.16: 'A working party report on: guidelines on materials requirements for carbon and low alloy steels for H ₂ S containing environments in oil and gas production.' The Institute of Materials, London, 1995. ISBN 0 901716 95 2.
   
   
8. ISO/FDIS 15156-2: 'Petroleum and natural gas industries - Materials for use in H ₂ S containing environments in oil and gas production - Part 2: Cracking resistant carbon and low alloy steels.' ISO Central Secretariat, Geneva, May 2003.
   
   
9. Coudreuse L, Zmudzinski C, Bocquet P and Cheviet A: 'Quenched and tempered steels for H ₂ S containing pressure vessels.' Paper presented at Second International Conference on Interaction of Steels with Hydrogen in Petroleum Industry Pressure Vessel and Pipeline Services, Vienna, Austria, 19-21 October 1994. Preprints Vol.1, pp 401-408. Materials Properties Council/TUV Osterreich, 1994.
   
   
10. Luxenburger G, Streisselberger A, Schwinn V and Demmerath A: 'Plates with homogeneous HIC-resistance.' Hydrocarbon Engineering May 1998, pp1-6.
    
    
11. Coudreuse L, Zmudzinksi C, Bocquet P, Gingell A and Cheviet A: 'Application of normalised and quenched and tempered pressure vessel steels as a function of sour service severity'. NACE Corrosion/2001, Houston, Texas, USA, 11-16 March 2001. Paper 01105.
    
    
12. Crolet J-L and Adam C: 'SOHIC without H ₂ S'. Materials Performance Vol.39, No.3, March 2000, pp86-90.
    
    
13. Schmitt G, Lohe G and Brukhoff W: 'Influence of corrosion products on the efficiency of inhibitors in sour gas service'. NACE Corrosion/87, San Francisco, March 1987. Paper 33.
    
    
14. Christensen C: 'SOHIC - so what?' NACE Corrosion/99. San Antonio, Texas, USA, 25-30 April 1999. Paper 434.
    
    
15. Materials Properties Council. 'Final report of the findings of the MPC Sponsored programme on the study of the susceptibility of steels to cracking in aqueous hydrogen charging environments'. March 1999.
    
    
16. Al-Anezi M A, Frankel G S and Agrawal A K: 'Susceptibility of conventional pressure vessel steel to hydrogen-induced cracking and stress-orientated hydrogen-induced cracking in hydrogen sulfide containing diglycolamine solutions', Corrosion, 55 (11) November 1999, pp 1101-1109.
    
    
17. Amano K, Kawabata F and Abe H: 'Mechanism of Stress Oriented HIC in high strength micro-alloyed steels'. Second International Conference on Interaction of Steels with Hydrogen in petroleum industry pressure vessel and pipeline service, Vienna, Austria, 19-21 October 1994. Preprints Vol 1, pp409-414. Materials Properties Council/TUV Osterreich, 1994.
    
    
18. Takahashi A, Ogawa H and Tamehiro H: 'Sulphide Stress Cracking at the softened heat affected zone in submerged arc welds of high strength line pipe steels'. International Conference on Pipeline Reliability, Calgary, Canada, 2-5 June 1992. Vol II, paper VI-7, pp 1-12. Gulf Publishing Company, Houston.
    
    
19. Iino M: 'The extension of hydrogen blister crack array in linepipe steels'. Metallurgical Transactions A Vol 9A, November 1978, pp 1581-1590.
    
    
20. Ikeda A and Kowaka M: 'Stress corrosion cracking of low and high-strength steels in wet hydrogen sulfide environment'. Chemical Economy and Engineering Review Vol 10, No.5-6 (No.117), May/June 1978, pp12-22.
    
    
21. Miyoshi E, Tonaka T, Terasaki F and Ikeda A: 'Hydrogen Induced Cracking of steels under wet hydrogen sulphide environment'. Petroleum Mechanical Engineering Conference, Tulsa, Oklahoma, USA, 21-25 September 1975. Paper 75-Pet-2 and Journal of Engineering for Industry, Trans.ASME Series B, Vol.98, No.11 1976 pp 1221-1230.
    
    
22. Bruckhoff W, Geier O, Hofbauer K, Schmitt G and Steinmetz D: 'Rupture of a sour gas line due to Stress Orientated Hydrogen Induced Cracking - Failure analyses, experimental results and corrosion prevention'. NACE Corrosion/85, Boston, Massachusetts, USA, 25-29 March 1985. Paper 389.
    
    
23. Iino M: 'Influence of sulphur content on the hydrogen induced fracture in linepipe steels'. Metallurgical Transactions A Vol.10A, November 1979, pp 1691-1697.
    
    
24. Gingell A D B and Garat X: 'Observations of damage modes as a function of microstructure during NACE TM0177-96 tensile testing of API 5L grade X60 linepipe steels'. NACE Corrosion/99. San Antonio, Texas, USA, 25-30 April 1999. Paper 600.
    
    
25. Ume K, Taira T, Hyodo T and Kobayashi Y: 'Initiation and propagation morphology of Sulphide Stress Corrosion Cracking at welds on linepipe steels.' NACE Corrosion/85, Boston, Massachusetts, 25-29 March 1985. Paper 240.
    
    
26. Blondeau R: 'Problems related to use of low alloy steels in H2S environments: industrial solutions'. Ironmaking and Steelmaking Vol.18, No.3, March 1991, pp 201-210.
    
    
27. Cayard M S and Kane R D: 'Large scale wet hydrogen sulfide cracking performance : Evaluation of metallurgical, mechanical and welding variables'. Corrosion, 53 (3), March 1997 pp 227 - 233.
    
    
28. Gianetto J A, Braid J E M and Bouker J T: 'Microstructural characterization of Hydrogen cracks in HAZ regions of selected pipeline steels.' Metals Technology Laboratories Report IIW Doc. No IX 1696.92 (MTL 92-06 (TR)). Canada Centre of Mineral and Energy Technology (CANMET), 1992.
    
    
29. Cialone H J and Williams D N: 'Sensitivity to Sulphide Stress Cracking at welds in line-pipe steels'. ASTM International Symposium on Environmentally Assisted Cracking: Science and Engineering, Bal Harbour, Florida, USA, 9-11 November 1987. ASTM Special Technical Publication STP 1049, pp152-168.
    
    
30. Kabayashi J, Ouchi H, Mabuchi H and Kikutake T: 'Evaluation of SSC initiation and propagation properties of high strength steels for pressure vessel'. Second International Conference on Interaction of Steels with Hydrogen in Petroleum Industry Pressure Vessel and Pipeline Service, Vienna, Austria, 19-21 October 1994. Preprints, Vol.1, pp143-154. Materials Properties Council/TUV Osterreich, 1994.
    
    
31. Fowler C, Haumann W and Koch F O: 'Influence of residual stresses to the SSCC resistance of SAW-pipe'. 3R International Vol.25, No.5, May 1986, pp.255-260.
    
    
32. Merrick R D and Bullen M L: 'Prevention of cracking in wet H ₂ S environments'. NACE Corrosion/89, New Orleans, Louisiana, 17-21 April 1989. Paper 269.
    
    
33. NACE Standard Test Method 'Laboratory Test Procedures for Evaluation of SOHIC Resistance of Plate Steels Used in Wet H ₂ S Service', TM0103-2003.
    
    
34. Siegmund G, Bruckhoff W, Schmitt G & Sadlowsky B 'SOHIC testing methods and their service relevance' NACE Corrosion 94 San Antonio, Texas, USA, 25-30 April 1999 Paper 433.
    
    
35. Fowler C M: 'The full ring test. A comprehensive proving test method for pipelines and associated welds'. Second International Conference on Interaction of Steels with Hydrogen in Petroleum Industry Pressure Vessel and Pipeline Service, Vienna, Austria, 19-21 October 1994. Preprints, Vol 1, pp 207-222. Materials Properties Council/TUV Osterreich, 1994.
    
    
36. Fowler C M and Bray J A: 'A test method to determine the susceptibility to cracking of linepipe steels in sour service'. Offshore Technology Report OTI 95 635, UK Health and Safety Executive, 1996.
    
    
37. Ikeda A: 'Progress on metallurgical investigation in HIC phenomena and development of HIC resistant steel'. Second International Conference on Interaction of Steels with Hydrogen in Petroleum Industry Pressure Vessel and Pipeline Service, Vienna, Austria, 19-21 October 1994 Preprints, Vol 1, pp 161-193. Materials Properties Council/TUV Osterreich, 1994.
    
    
38. Takahashi A, Hara T, Ogawa H: 'Susceptibility to stress oriented HIC of an X65 UOE linepipe evaluated by full ring test and small scale tests' Ibid, Vol 2, pp 699-705.
    
    
39. TWI archive material relating to the Aramco Gart pipeline failure.
    
    
40. Private Communication.
    
    
41. Hay M G Private Communication
    
    
42. Hay M G & Rider D W 'Integrity management of a HIC-damaged pipeline and refinery pressure vessel through hydrogen permeation measurements' NACE Corrosion/98, San Diego, California March 22-27 1998, Paper 395.
    
    
43. McHenry H I, Shives T R, Read D T, McColskey J D, Brady C H & Purtscher, P T 'Examination of a pressure vessel that ruptured at the Chicago refinery of the Union Oil Company on July 23 1984' Occupational Safety & Health Administration, US Department of Labor March 1986.
    
    
44. Gazam S Y & El-Amari A 'Failure of a gas /condensate line' Materials Performance, 31(10) October 1992 pp55-56.
    
    
45. Jones S, Private Communication.
    
    
46. NACE technical committee special report: 'SSC resistance of pipeline welds'. Materials Performance, 32(1) January 1993, pp58-64.
    
    
47. 'Accurate assessment of material susceptibility to SOHIC for line pipe and pressure vessel steels' Joint Industry Research Project, Bodycote and TWI.